What Targeted Provisioning Can—and Cannot—Do

Colombia’s maturity-specific provisioning reform answers a narrower question than “did credit tighten?” It shows how a prudential rule can raise the buffer attached to long-maturity consumer loans, leave average borrower-facing terms broadly unchanged, and still tighten the longest loans at smaller institutions. The useful final judgment is therefore conditional: the reform appears to have strengthened preparedness for a defined risk, but it was not designed—and the evidence does not show it—to control the entire credit cycle or guarantee equal access across lenders.

One chain, two levels of result

The case can be reconstructed as one connected chain. A long repayment horizon leaves a lender exposed to more possible changes in a borrower’s income, the economy, collateral values, and repayment conditions. Colombia’s expected-loss framework represents that exposure through probability of default, exposure at default, loss given default, a maturity adjustment, and the additional K factor. For new covered loans, K is 1 up to 72 months, 1.1 above 72 and up to 108 months, and 1.4 above 108 months. External Circular 026 therefore raised the expected-loss cost in discrete maturity bands rather than applying one uniform increase to all consumer credit.

That change is the policy shock. It does not dictate a borrower’s interest rate or a lender’s approval decision. It first changes what the institution must recognize against potential loss. The institution then chooses how to manage the added cost. It can absorb the expense, pass part of it through pricing, reduce the principal, lower the loan-to-value ratio, prefer a shorter maturity, or alter which applications it accepts. The same rule can thus create a common balance-sheet pressure and different contract outcomes.

The study’s design follows that chain. It uses granular supervisory records of newly originated covered consumer loans during 2022–2024, then compares maturity groups after making them more alike on observable loan, borrower, lender, and time characteristics. The two treatment tiers are loans above 72 and up to 108 months and loans above 108 months; the control group contains loans at or below 72 months. Exact matching and weighted difference-in-differences are meant to separate the additional policy effect from the fact that long loans were already different products for different borrowers and lenders.

The first result is the most direct one: provision coverage rose. Under the strictest matching specification, the estimated increase was about 0.166 percentage points at the 72-month threshold and about 0.86 percentage points at the 108-month threshold. The positive estimates persist across the matching specifications. The larger movement in the longest tier is consistent with its larger additional factor. In the language of policy objectives, the rule strengthened the forward-looking buffer for the affected new loans.

That finding should not be enlarged into a claim about realized defaults or household safety. A provision is preparation for a possible loss. A higher coverage ratio can make an institution better able to absorb deterioration, but it does not prove that a borrower will default, that the institution will ultimately lose less, or that the household’s debt burden has become more manageable. The result is an ex ante resilience effect, not a complete forecast of the credit cycle.

The second result concerns the contract terms visible in the aggregate matched sample. The study finds no statistically distinguishable change in interest rates, loan amounts, or LTV ratios at either threshold under the stricter specifications. Higher provisions were not broadly passed through as higher prices, smaller loans, or greater collateral requirements. This is not the same as saying that no lender changed anything.